← Back to the model

Status & Pilot Readiness

Last updated 2026-10-20 11:35 UTC-05:00

Ready to pilot: from Nov 9

Ready for the joint staffing pilot: Dec 1 – Jan 15

A pilot date is when a part is ready to be piloted, not a booked slot, because piloting waits on incoming projects. Parts get ready on their own, manually if needed; only the staffing harness has to wait for the others.

Mapping looks back at 2–3 running and 3–4 recent projects. A part moves on at 75–80% coverage. Dates are estimates and never move a part.

What's happening today

Latest model change

  • Manual staffing run: audit step added from the Claude CoWork run
  • Record the manual run findings as open questions; affinity goes on the decisions list
  • Fill math gaps: defined terms, domain units, scope blocks for every capability, manual staffing run
  • Glossary: drop confidence callouts; add a nav hide/show toggle
  • Correction versus variation: derivation field, override records, checkOverrides, Kind column on the glossary

What's happening tomorrow

  • Capabilities and levels work. It includes self-assessment of people's execution levels, which also checks the levels themselves.
  • Formulas for surface area and required level. Surface area is drafted, and required level waits on how stake is set for each risk shape.
  • Settling the terms the model rests on. Instances, concurrency and allocation as separate attributes.

Where each part is, and when it could moveA part moves right only after real work holds it up, and moves back left if real work breaks it. A group sits in the column of its least tested part.

Each bar covers its likely window.

Thinking timeWe have written it down and talked it through. We have not tried it on real work yet. MappingLooks back. Does this part explain real work we have already done, in running engagements and recently finished projects? PilotLooks forward. We use this part on an upcoming real project and compare the result later. A manual or spreadsheet run counts, and a pilot should be the smallest chunk that gives real feedback. Dates show when a part is ready to pilot, since the pilot itself waits on an incoming project.
Part of the model
Oct 2026Nov 2026Dec 2026Jan 2027Today
Pilot not defined yet
Stages 0–1
Piloted with the staffing harness
Piloted with the staffing harness
Piloted with the staffing harness
Pilots on its own

Ahead of the estimate

  • Sizing formulas and defined terms landed ahead of their thinking windows.
  • A manual staffing run, written down step by step, ahead of the harness build.
  • Human overrides split into corrections and variations before any pilot.
  • No part has moved column; mapping against real work has not started.

Why this isn't a sequential roadmap

The parts depend on each other in a loop, so none of them finishes first. An arrow runs from the part that is needed to the part that needs it, a dot shows the column a part sits in, and a dashed arrow is feedback that does not exist yet. Hover a letter to see why.

ABCDEFG Capability Levels Roles & Titles Staffing harness Sizing & Staffing Rules AI-Native SDLC Operating View Capability Levels → Roles & Titles. A title is a bundle of capabilities at levels, so it cannot settle before the levels do.Roles & Titles → Staffing harness. The briefing stops at the qualified pool, and resourcing reads titles to fill it.Sizing & Staffing Rules → Staffing harness. The harness needs to know how many instances of each capability the pilot's work calls for.AI-Native SDLC → Sizing & Staffing Rules. What counts as one unit depends on how the work is cut into stages.AI-Native SDLC → Operating View. Risk shapes run hot at particular stages, so they follow the lifecycle's stage map.Operating View → Capability Levels. Risk shapes and dials say how hard a capability runs, which is what a level has to be able to describe.Staffing harness → Capability Levels. Staffing runs should show where a level was set wrong. This feedback does not exist until the harness has run once. Not real yet.

What's done

The model

  • Six domains with 26 capabilities, each on the L1 to L4 execution scale.
  • Five titles, with every capability owned by exactly one.
  • Fifteen definitions, each with the confusions it rules out.
  • Scope Units, with a default unit and leverage for all 26 capabilities.

How work runs

  • The AI-Native SDLC, seven stages plus its agentic mode.
  • Eight risk shapes, the intensity dial, and five seams.
  • Four doctrines, including New Discovery and what happens when work changes.
  • The build and QA seat rule.

Infrastructure

  • YAML as the source of truth, with a schema and validator for every entity.
  • A generated site, rebuilt from the YAML.
  • A read-only MCP server that agents use to read the model.
  • Visibility tiers and a private overlay path for internal data.

Harness ideation

  • Staffing and capability mapping: designed; ledger row shape defined.
  • ESOA: designed; refresh trigger open.
  • Evidence Work: drafted; may trail the pilot.

Decisions we need

  • Define what Responsible means for Brian on items within the ADLC.
  • There's a boundary and we're not sure where it lands. The original intention for the entire model is that it isn't scoped just to the consulting group; it includes all ICs who participate in the ADLC, regardless of their services or consultant status. The question is: Should it be?
  • How many engagements we need to map against, and how long we will wait for them before mapping is skipped.
  • Whether capability affinity, meaning pairs that are cheap to co-hold, deserves a field of its own, or stays a judgment call. Segregation covers pairs that must never share a seat and allocation covers the arithmetic, but nothing records that acceptance proving and demonstration evidence review cost little extra when the same person is building, while validation testing and signal design are separate disciplines that nothing segregates.

What moves the range

Pushes it later

  • A large re-cut of the levels.
  • Thin delivery data from the mapping engagements.
  • More than one round on scope units.
  • Holidays.

Pulls it earlier

  • A pilot scoped to a few capabilities.
  • The ledger schema landing next week.
  • Letting the Evidence Work harness trail the pilot.
  • Running the staffing harness pilot manually, ahead of the agent harness build.

Glossary: the terms and rules behind sizing and staffing →